Method for correcting inhomogeneity of the static magnetic field particularly of the static magnetic field generated by the magnetic structure of a machine for acquiring nuclear magnetic resonance images and mri system for carrying out such method

Abstract

Shimming a magnetic field includes measuring the magnetic field and sampling it in a plurality of locations; defining a grid for positioning correction elements; calculating the position and magnitude parameters of one or more correction elements to obtain predetermined target values of the field characteristics, wherein an algorithm has been trained by a database of known cases in which each record links a certain initial magnetic field of a magnetic structure to the pattern of correction elements on a positioning grid for these correction elements in the magnet structure, thereby delivering as an output a pattern of correction elements on the positioning grid, which contributions to the magnetic field of the magnet structure generate a magnetic field which at least best approximates or meets the predetermined target values of the field characteristics.

1 - 15 . (canceled) 
     
     
         16 . A system for carrying out shimming of a magnetic structure, the system comprising:
 a sensor unit for measuring a magnetic field of the magnetic structure;   a supporting structure of the sensor unit which can be displaced along three spatial coordinates for positioning the sensor unit at different spatial positions;   a data collection unit for receiving the magnetic field measures of the sensor unit at each spatial position and comprising a memory for saving data pairs relating to magnetic field strength and spatial position at which the magnetic field strength has been measured;   a processing unit comprising a memory in which a machine learning algorithm model is saved and a working memory in which said machine learning model is loaded for execution;   a data output comprising a display or a printer to output a list of positioning coordinates on a positioning grid of one or more magnetic correction elements and a corresponding magnetic charge of the correction elements or of an image of a positioning grid of the correction elements on which grid the correction elements are positioned according to the calculated distribution and are combined with the visual indication of the corresponding magnetic charge.   
     
     
         17 . A system for carrying out shimming of a magnetic structure, the system comprising:
 a sensor unit for measuring a magnetic field of the magnetic structure;   a supporting structure of the sensor unit which can be displaced along three spatial coordinates for positioning the sensor unit at different spatial positions;   a data collection unit for receiving the magnetic field measures of the sensor unit at each spatial position and comprising a memory for saving the data pairs relating to magnetic field strength and spatial position at which the magnetic field strength has been measured;   a processing unit comprising a memory in which a machine learning algorithm model is saved and a working memory in which said machine learning model is loaded for execution;   an automatic pic and place device comprising a robotic arm for the correction elements, which is provided in combination with a magazine of differently magnetically charged correction elements and a control unit of the robotic arm which controls the pic and place operations of the magnetic correction elements on a positioning grid according to a computed distribution, and which control unit receives coordinates of the position of each of the magnetic correction elements and information of which kind of correction element is to be put in place at a certain coordinate on the grid from the processing unit and generates commands to drive the robotic arm.   
     
     
         18 . The system according to  claim 16  in which the supporting structure of the sensor unit for positioning the sensor unit at different spatial positions is a pick and place device carrying the sensor unit for measuring the magnetic field. 
     
     
         19 . The system according to  claim 16  wherein the magnetic structure is an MRI apparatus. 
     
     
         20 . The system according to  claim 16  wherein in the machine learning algorithm is an algorithm trained by a database of known cases in which each record links a certain initial magnetic field of a magnetic structure to a pattern of correction elements on a positioning grid for these correction elements in the magnet structure, thereby delivering as an output a pattern of correction elements on the positioning grid which contributions to the magnetic field of the magnet structure generate a magnetic field which at least approximates or meets the predetermined target values of the field characteristics, and wherein a distribution of correction elements of a certain predetermined positioning grid is combined with a training process which provides for a selection of specific records from the database of known cases which allows to generate a training database providing training of the machine learning algorithm;
 wherein the training process also provides a step for reducing a number of correction elements on a grid by individuating the correction elements of the distribution of correction elements which can be substituted by a smaller number of correction elements at different positions and having a different magnetic charge. 
 
     
     
         21 . The system according to  claim 16  in which the machine learning algorithm is set to predict the distribution of correction elements on a grid for positioning the correction elements in a magnet structure which minimizes a difference between a theoretically wanted magnetic field and a measured magnetic field. 
     
     
         22 . The system according to  claim 16  in which the machine learning algorithm is a predictive algorithm trained on the database of known cases and which uses as an input sample data of the magnetic field effectively measured and as an output the pattern of the correction elements on the grid. 
     
     
         23 . The system according to  claim 16  in which the machine learning algorithm is trained by a training database comprising records relating to the following data of each one of a population of known cases:
 the measured data of the magnetic field in its initial non optimized, i.e., shimmed condition, at measuring points distributed according to a predetermined distribution in a predetermined volume of space; 
 optionally the size or the position in relation to the magnet structure or the shape of the boundary surface of the volume of space; 
 the chosen kind of grid for positioning the correction elements of the magnetic field; 
 the correction elements determined by the successfully executed process of optimization or shimming, each one of the correction elements being characterized by one or more of the following parameters: position on the grid, magnetic charge, i.e., strength of the magnetic field generated by the correction elements, and polarity of the magnetic field. 
 
     
     
         24 . The system according to  claim 16  in which the machine learning algorithm is chosen as one or a combination of the following algorithms:
 a genetic algorithm in which the distribution of correction elements on a certain grid which cancels or reduces the inhomogeneities is calculated by combining known distributions in a database of known cases in order to generate a distribution of correction elements minimizing the inhomogeneities, the evolutionary process for combining the known distributions in order to generate the next generations is guided by an inhomogeneity pattern of the magnetic field corrected through the distribution of correction elements resulting from the theoretical calculation of the magnetic field at predetermined measuring positions on a grid of measuring points defined in a volume permeated by the magnetic field; 
 a predictive algorithm consisting in a neural network in which nodes represents the grid of the correction elements and a matrix of the weights which comprises a weight applied to each activation function of each node is the output determining the magnetic field of the correction element at a certain position of the grid; 
 a classification algorithm; or 
 a combination of two or more of the above algorithms. 
 
     
     
         25 . The system according to  claim 16 , in which in the memory of the processing unit there is loaded with a control program which comprises the instructions for the processing unit which instructions when executed by the processing unit make the system able to carry out the following steps:
 a) measuring the magnetic field and sampling it in a plurality of locations, with a predetermined space distribution within a predetermined volume of space permeated by the magnetic field and delimited by a surface boundary or also on the surface boundary;   b) defining a grid for positioning correction elements based on a magnet structure and on the a correlation thereof with a magnetic field structure;   c) calculating position and magnitude parameters of one or more correction elements to obtain predetermined target values of magnetic field characteristics, in which the step c) is carried out by a machine learning algorithm or combinations thereof,   which algorithm has been trained by a database of known cases in which each record links a certain initial magnetic field of a magnetic structure to a pattern of correction elements on a positioning grid for these correction elements in the magnet structure, thereby delivering as an output a pattern of correction elements on the positioning grid which contributions to the magnetic field of the magnet structure generate a magnetic field which at least approximates or meets the predetermined target values of the field characteristics   wherein the algorithm is executed by   individuating a symmetry of the magnet structure or of the grid and dividing the magnet structure or the grid or the magnetic field into sectors according to the symmetry;   determining the one or more correction elements and their positions on the grid within the one first sector;   projecting the one or more correction elements and their positions on the grid from the first one sector to each of the further sectors according to the symmetry; and   carrying out the above steps for each repetition/iteration step/steps of the method when one or more of such repetition/iteration steps are provided   and in in which the steps a) to c) are repeated two or more times and in which at each repetition/iteration the magnetic field measured is the field generated by the magnetic structure in combination with the magnetic field generated by the correction elements determined and placed as the result of the previous repetitions/iterations,   the number of repetitions/iteration in a sequence of repetitions/iterations is terminated according to one criteria or to a combination of criteria of the following list:   the measured inhomogeneity of the shimmed magnetic field is below a predetermined maximum inhomogeneity threshold,   the number of repetitions has reached a predetermined maximum number of repetitions,   the rate of reduction of the inhomogeneity at each step falls below a minimum rate value, i.e. each new repetition does not provide for a relevant further reduction of the inhomogeneity in relation to the previous ones and/or the difference of the effectively generated magnetic field in relation to a desired theoretically determined target field are below a certain maximum value of a parameter or a combination of parameters describing the difference between the magnetic fields.   
     
     
         26 . The system according to  claim 17  in which the supporting structure of the sensor unit for positioning the sensor unit at different spatial positions is a pick and place device carrying the sensor unit for measuring the magnetic field. 
     
     
         27 . The system according to  claim 17  wherein the magnetic structure is an MRI apparatus. 
     
     
         28 . The system according to  claim 17  wherein in the machine learning algorithm is an algorithm trained by a database of known cases in which each record links a certain initial magnetic field of a magnetic structure to a pattern of correction elements on a positioning grid for these correction elements in the magnet structure, thereby delivering as an output a pattern of correction elements on the positioning grid which contributions to the magnetic field of the magnet structure generate a magnetic field which at least approximates or meets the predetermined target values of the field characteristics, and wherein a distribution of correction elements of a certain predetermined positioning grid is combined with a training process which provides for a selection of specific records from the database of known cases which allows to generate a training database providing training of the machine learning algorithm;
 wherein the training process also provides a step for reducing a number of correction elements on a grid by individuating the correction elements of the distribution of correction elements which can be substituted by a smaller number of correction elements at different positions and having a different magnetic charge. 
 
     
     
         29 . The system according to  claim 17  in which the machine learning algorithm is set to predict the distribution of correction elements on a grid for positioning the correction elements in a magnet structure which minimizes a difference between a theoretically wanted magnetic field and a measured magnetic field. 
     
     
         30 . The system according to  claim 17  in which the machine learning algorithm is a predictive algorithm trained on the database of known cases and which uses as an input sample data of the magnetic field effectively measured and as an output the pattern of the correction elements on the grid. 
     
     
         31 . The system according to  claim 17  in which the machine learning algorithm is trained by a training database comprising records relating to the following data of each one of a population of known cases:
 the measured data of the magnetic field in its initial non optimized, i.e., shimmed condition, at measuring points distributed according to a predetermined distribution in a predetermined volume of space; 
 optionally the size or the position in relation to the magnet structure or the shape of the boundary surface of the volume of space; 
 the chosen kind of grid for positioning the correction elements of the magnetic field; 
 the correction elements determined by the successfully executed process of optimization or shimming, each one of the correction elements being characterized by one or more of the following parameters: position on the grid, magnetic charge, i.e., strength of the magnetic field generated by the correction elements, and polarity of the magnetic field. 
 
     
     
         32 . The system according to  claim 17  in which the machine learning algorithm is chosen as one or a combination of the following algorithms:
 a genetic algorithm in which the distribution of correction elements on a certain grid which cancels or reduces the inhomogeneities is calculated by combining known distributions in a database of known cases in order to generate a distribution of correction elements minimizing the inhomogeneities, the evolutionary process for combining the known distributions in order to generate the next generations is guided by an inhomogeneity pattern of the magnetic field corrected through the distribution of correction elements resulting from the theoretical calculation of the magnetic field at predetermined measuring positions on a grid of measuring points defined in a volume permeated by the magnetic field; 
 a predictive algorithm consisting in a neural network in which nodes represents the grid of the correction elements and a matrix of the weights which comprises a weight applied to each activation function of each node is the output determining the magnetic field of the correction element at a certain position of the grid; 
 a classification algorithm; or 
 a combination of two or more of the above algorithms. 
 
     
     
         33 . The system according to  claim 17 , in which in the memory of the processing unit there is loaded with a control program which comprises the instructions for the processing unit which instructions when executed by the processing unit make the system able to carry out the following steps:
 a) measuring the magnetic field and sampling it in a plurality of locations, with a predetermined space distribution within a predetermined volume of space permeated by the magnetic field and delimited by a surface boundary or also on the surface boundary;   b) defining a grid for positioning correction elements based on a magnet structure and on the a correlation thereof with a magnetic field structure;   c) calculating position and magnitude parameters of one or more correction elements to obtain predetermined target values of magnetic field characteristics, in which the step c) is carried out by a machine learning algorithm or combinations thereof,   which algorithm has been trained by a database of known cases in which each record links a certain initial magnetic field of a magnetic structure to a pattern of correction elements on a positioning grid for these correction elements in the magnet structure, thereby delivering as an output a pattern of correction elements on the positioning grid which contributions to the magnetic field of the magnet structure generate a magnetic field which at least approximates or meets the predetermined target values of the field characteristics   wherein the algorithm is executed by   individuating a symmetry of the magnet structure or of the grid and dividing the magnet structure or the grid or the magnetic field into sectors according to the symmetry;   determining the one or more correction elements and their positions on the grid within the one first sector;   projecting the one or more correction elements and their positions on the grid from the first one sector to each of the further sectors according to the symmetry; and   carrying out the above steps for each repetition/iteration step/steps of the method when one or more of such repetition/iteration steps are provided   and in in which the steps a) to c) are repeated two or more times and in which at each repetition/iteration the magnetic field measured is the field generated by the magnetic structure in combination with the magnetic field generated by the correction elements determined and placed as the result of the previous repetitions/iterations,   the number of repetitions/iteration in a sequence of repetitions/iterations is terminated according to one criteria or to a combination of criteria of the following list:   the measured inhomogeneity of the shimmed magnetic field is below a predetermined maximum inhomogeneity threshold,   the number of repetitions has reached a predetermined maximum number of repetitions,   the rate of reduction of the inhomogeneity at each step falls below a minimum rate value, i.e. each new repetition does not provide for a relevant further reduction of the inhomogeneity in relation to the previous ones and/or the difference of the effectively generated magnetic field in relation to a desired theoretically determined target field are below a certain maximum value of a parameter or a combination of parameters describing the difference between the magnetic fields.